The Center for Space Standards & Innovation
In order to reduce latency and improve accuracy, CelesTrak now offers supplemental two-line element sets (TLEs) derived directly from owner/operator-supplied orbital data. Until now, all TLEs provided to the public have been derived from radar and optical observations of the US Space Surveillance Network (SSN). This process can cause problems, particularly for deep-space objects (orbital periods greater than 225 minutes) which rely on optical observations. Limited geographic distribution of these optical sites, together with unpredictable weather conditions, can result in long intervals between observations, resulting in delays in cataloging objects, old and inaccurate data, and even difficulty in recovering objects.
While this uncooperative approach to tracking space objects is required for a large portion of the current satellite catalog, many of these objects are operational payloads which are routinely tracked by their owners or operators. Some of this orbital data, such as for the GPS constellation, is publicly available via the Internet. The goal of this new service is to take that data and generate TLEs as part of the normal data update processes on CelesTrak.
Each day, CelesTrak checks known sources of publicly available orbital data and produces TLEs from that data using Satellite Tool Kit (STK). As an example, for the GPS constellation, the latest GPS almanac provided by the 2nd Space Operations Squadron and posted in the GPS Data section on CelesTrak is propagated in accordance with IS-GPS-200F to produce ephemerides for the coming day. Those ephemerides are used as 'observations' to fit a TLE using the SGP4 orbital model in STK and described in detail in the paper "Revisiting Spacetrack Report #3."
As a result, for the GPS constellation, where TLEs are generally released one to two days after they are generated, the uncertainty drops from several kilometers to several hundred meters (approximately an order of magnitude). For a detailed analysis of the accuracy of GPS almanacs and GPS TLEs, see the paper "Validation of SGP4 and IS-GPS-200D Against GPS Precision Ephemerides." A more detailed analysis of the improved accuracy of these supplemental TLEs will be provided in an upcoming paper. For now, test cases for the GPS constellation on 2007 Dec 31 and the GLONASS constellation on 2008 Jan 23 are provided below.
Each supplemental TLE can be distinguished from regular NORAD TLEs by looking at the Classification field on Line 1, Column 8. TLEs provided by Space Track will have a "U" in this field (for Unclassified). Supplemental TLEs will have a "C" in this field (to identify the source as CelesTrak). GPS supplemental TLEs will use the GPS Week and day of that week as the Element Set Number (e.g., GPS Week 435, TOA 589824 yields Element Set Number 4356). Supplemental TLEs for other satellites will use similar methods to set the Element Set Number to allow users to track the TLE back to the source data.
For the Intelsat data, where both nominal ephemeris data is provided along with post-maneuver data, both TLEs are generated. That allows the user to use the nominal TLE up to the time of the maneuver and then switch to the post-maneuver TLE after the time of the maneuver. The post-maneuver TLEs are marked with [PM] at the end of the satellite name (Line 0).
In addition, information on the accuracy of the SGP4 fit for each satellite is provided for each set of TLE data, with a link to a file showing the final calculated RMS and the number of iterations required to meet the convergence criteria.
We are working to add additional supplemental TLEs, as we become aware of other publicly available orbital data sources.
Other satellite owner/operators wishing to take advantage of this free service should feel free to contact me about how to obtain their orbital data.
As suggested above, there are several advantages to this new approach to providing TLE data:
To demonstrate the improved accuracy available from these supplemental TLEs, a comparison of the TLEs released by Space Track at 1300 UTC on 2007 Dec 31 was made to the supplemental TLEs generated from the SEM almanac released that same morning at 1515 UTC. The comparison was made to the NGA GPS Satellite Precise Ephemeris for 2007 Dec 31, made available on the NGA GPS Satellite Precise Ephemeris page on 2008 Jan 2, which is accurate to the centimeter level.
It should be noted that the Space Track TLEs used here have epochs ranging from Day 361.44451533 to Day 363.82128857, which means they must be propagated forward anywhere from 1.7 to 4.1 days to the time they were released on 2007 Dec 31 (Day 365.54166667). The GPS almanac, however, used a Time of Availability (TOA) of 319488 seconds on GPS Week 436, which corresponds to 2008 Jan 2 at 16:44:34 UTC (several days in the future). That means the supplemental TLEs, which are generated with an epoch matching that TOA are propagated backward 2.16 days to the same time.
Figure 1 shows a comparison of the results for PRNs 1 through 6 and 8 through 32 (PRN07 is not yet part of the GPS constellation). The red curves show the error in the Space Track TLEs when compared to the NGA Precise Ephemeris on 2007 Dec 31 and the blue curves show the error in the supplemental TLEs. It should be obvious from the plot that the supplemental TLEs are far more accurate over this interval than the corresponding Space Track TLEs. The 31 Space Track TLEs show an average error of 7.54 km over this period with a maximum error (for PRN15) of 32.45 km compared to an average error for the supplemental TLEs of 0.87 km and a maximum error of 2.37 km, yielding approximately an order of magnitude improvement in the error.
Figure 1. Comparison of NORAD and Supplemental TLEs to NGA Precise Ephemerides
A comparison of the TLEs released by Space Track at 1300 UTC on 2008 Jan 25 was made to the supplemental TLEs generated from the GLONASS Rapid Precise Ephemerides released that same morning at 1311 UTC, covering the period from 2008 Jan 23 23:59:46.000 UTC (0000 GPS Time) to 2008 Jan 24 23:44:46.000 UTC. The comparison was made to the Final Precise Ephemerides for the day both data sets were released.
It should be noted that the Space Track TLEs used here have epochs ranging from Day 023.88315278 to Day 025.47923451 at the time they were released on 2008 Jan 25 (Day 025.54166667). The GLONASS Rapid Precise Ephemerides used to generate the supplemental TLEs, however, cover the period of Day 023.99983796 to Day 024.98942130.
Figure 2 shows a comparison of the results for GLONASS Slots 1, 4, 6, 7, 8, 10, 11, 14, 15, 17, 19, 20, 23, and 24. The red curves show the error in the Space Track TLEs when compared to the GLONASS Final Precise Ephemeris on 2008 Jan 25 and the blue curves show the error in the supplemental TLEs. It should be obvious from the plot that the supplemental TLEs are far more accurate over this interval than the corresponding Space Track TLEs. The 14 Space Track TLEs show an average error of 3.30 km over this period with a maximum error (for Slot 7) of 9.39 km compared to an average error for the supplemental TLEs of 0.20 km and a maximum error of 0.54 km, yielding over an order of magnitude improvement in the error.
Figure 2. Comparison of NORAD and Supplemental TLEs to GLONASS Final Precise Ephemerides
Similar results should be expected for geostationary satellites, although a demonstration of that performance will have to await further analysis.
Dr. T.S. Kelso
Follow CelesTrak on Twitter @TSKelso
Last updated: 2014 May 17 02:38:37 UTC
Accessed 98,502 times since 2000 December 16
Current system time: 2014 August 1 09:49:06 UTC